Acknowledgements Table of Contents Sections 1-3 Section 4 Sections 5-6

5. Products of the Initiative

5.1 Products of the Specialist Meeting

The primary product of the Specialist Meeting is the research agenda. Other items have been identified as potential products of the Initiative.

5.1.1 Bibliography

While there was some discussion that the bibliography be organized within the framework suggested by Lew Hopkins' summary contexts, the group preferred that it be based on keywords. Participants will be sent the list of appropriate keywords and asked to code their own position paper references and other relevant references before submitting them to NCGIA for incorporation into a master bibliography to be published as a technical report

5.1.2 Closing conference

The group strongly recommended that the initiative leaders begin working on the establishment of a "I-17 closing conference" to be held in about 2 years time. This meeting may be similar in organization to a NATO ARW. Papers offered for presentation at the meeting will be referred in their entirety and collected into a formal published book. The conference is to be presented now as a challenge to meeting participants to encourage them to move forward on research issues identified here and to be prepared to demonstrate to their colleagues progress they have made since this meeting.

5.1.3 WWW homepage for I-17

The group requested that the homepage for I-17 be maintained and updated to provide for continued collaboration between meeting participants. The homepage may provide a number of services including:

5.1.4 Book and journal articles

The group was not supportive of the idea of a book arising immediately from this meeting but agreed that discussion between meeting participants to develop joint papers should be encourage.

5.1.5 Conference sessions

Several conferences were identified as likely places for papers on CSDM. Participants will cooperate through the I-17 WWW homepage to develop potential paper topics. Before the end of the specialist meeting, a special session was organized for the next International Conference on Integrating GIS and Environmental Model.

5.1.6 Critique of the meeting as a collaborative effort

It was suggested that one or more participants of the meeting with relevant experience in the area be enlisted to write a very brief critique of the meeting as a collaborative effort. Participants felt that such a review would be an interesting byproduct of the Specialist Meeting forum.

5.2 Papers prepared to date

5.2.1 Refereed journals

Armstrong, M.P. (1994) Requirements for the development of GIS-based group decision support systems. Journal of the American Society for Information Science 45(9): 669-677.

Armstrong, M.P. (forthcoming) Is there a role for high performance computing in GIS? Journal of the Urban and Regional Information Systems Association.

Densham, P.J. and G. Rushton (forthcoming) Providing spatial decision support for rural service facilities that require a minimum workload. Environment and Planning B.

5.2.2 Book chapters

Armstrong, M.P. and P.J. Densham (1995) A conceptual framework for improving human-computer interaction in locational decision-making. In Nyerges, T., D. Mark, R. Laurini, and M. Egenhofer (eds.) Cognitive Aspects of Human-Computer Interaction for Geographic Information Systems. Kluwer, Dordrecht: 343-354.

Densham, P.J. and M.P. Armstrong (1995) Human-computer interaction considerations for visual-interactive locational analysis. In Nyerges, T., D. Mark, R. Laurini, and M. Egenhofer (eds.) Cognitive Aspects of Human-Computer Interaction for Geographic Information Systems. Kluwer, Dordrecht: 179-196.

Densham, P.J. (forthcoming) Visual interactive locational analysis. In Longley, P., and M. Batty (eds.) Spatial Analysis: Modeling in a GIS environment. GeoInformation International, Cambridge

5.2.3 Conference proceedings

Armstrong, M.P. and P.J. Densham (in press) Toward the development of a conceptual framework for GIS-based collaborative spatial decision-making. Proceedings of the Second ACM Workshop on Advances in Geographic Information Systems, Gaithersberg, MD

Armstrong, M.P. and P.J. Densham (1995) Cartographic support for collaborative spatial decision-making. Proceedings of the 12th International Symposium on Automated Cartography (Auto-Carto 12), Bethesda, MD: 49-58.

Densham, P.J. and M.P. Armstrong (1994) A heterogeneous processing approach to spatial decision support systems. In Waugh, T.C., and R.G. Healey (eds.) Advances in GIS Research: Proceedings of the Sixth International Symposium on Spatial Data Handling, Volume 1. Taylor and Francis, London: 29-45.


During the course of the specialist meeting, participants developed a research agenda for CSDM which centers around 2 major themes: tool development and tool use. Research questions that relate to tool development can be grouped into those concerned with assessing and defining the tool requirements of individuals and groups, those that seek to exploit developments in cognate fields, and those that focus on the peculiarly spatial aspects of CSDM. In the case of tool use, research questions can be grouped into those that examine representation, those that seek to assess the effectiveness of CSDM software, and those that are concerned with the roles of users and mediators during CSDM and how they relate to different forms of CSDM software.

One of the outcomes of the specialist meeting is that a cadre of researchers have discussed the impediments to the widespread adoption of CSDM and have developed a common understanding of the magnitude and relative importance of these impediments. This shared understanding provides a starting point for research under the aegis of the Initiative. Many of the participants were working on parts of this agenda before the specialist meeting, others have indicated that they will adopt elements of it in their own research. A WWW server is planned to help these researchers coordinate their work and to be informed of what others are doing.

It is important to note that the formal termination of the initiative (currently planned for the summer of 1997) will not signal the end of research on CSDM. Rather, the research carried out during the life-span of the initiative will further refine the research agenda and make it accessible to a wider research community.

6.1 Related research activities at University of California at Santa Barbara

At UCSB, a small working group has been formed to continue work on topics related to this research initiative. This working group has defined six major research areas in Collaborative Spatial Decision Making resulting from discussions at this meeting. Most of these research areas are not unique to the spatial domain, but their solutions in the spatial domain require modification of existing models and development of new models and model interfaces.

6.1.1 Assess the usefulness of existing representations of spatial information for representing the spatial aspects of the interests of participants in multiparty decision making

Using a spatial decision support system to model and analyze spatial problems requires an adequate representation of the objectives and interests of the participants of the problem. This requires a sophisticated understanding of the geographical conceptions of the problem that are inherent in participants' interests. While one representation may be appropriate for one group and their interests, it may not adequately represent others. If the representations of the interests that are used in various models or presentations of information are not consistent with all participants' individual conceptions and across the decision space, then the results of models and decision support systems will not contribute to resolving disputes or producing collaborative decisions. Research is needed to identify typical spatial conceptualizations of problems for classes of spatial problems and for typical stances in these problems. Evaluations of the effectiveness of different existing methods of representing these conceptualizations can provide useful input to the design of spatial decision support systems and models for collaborative decision making.

6.1.2 Modeling with multiple data sets, multiple models, and multiple problem representations

In a computing environment designed to support collaborative decision making between several groups, often there is not complete agreement upon the data set to be used and the model to be employed. Thus it may be necessary to apply a model to any of a number of different data sets or to use any data set in all models. Thus a computing environment must be available to support multiple models and data sets, plus an interface which can aid in comparing alternatives, measuring differences between them, and presenting/viewing such alternatives.

For a spatial CSDM example, consider a situation in which one group in a decision process would like to use a median location model to locate ambulance stations in an urban area, while another group insists on the use of a maximal covering problem. While both groups agree to use the same data set, two different models will be employed. In order to communicate between groups, one model's output (say sites and weighted distance) needs to be compared in terms of the other model's objective (coverage within some distance standard).

Although this sounds simple, negotiation would require generating and presenting compromise solutions. To do that would require one of two techniques: 1) a multiobjective model which supports both objectives, or 2) a methodological bridge which can systematically integrate two independent models with weights and structural conditions which can be used to identify compromise solutions. The first approach requires that the integrated model exists in the first place and that all integration is done in advance and has been anticipated. The second approach has never been attempted or theoretically scoped out.

6.1.3 Generating alternatives

A major need in the support of collaboration in spatial decision making is the capability to generate alternatives that achieve specific objectives or have specific spatial qualities. Frequently, however, decision makers are not able to specify all their objectives completely, thus some objectives remain hidden or private. Brill, Hopkins and others have argued that when hidden objectives are exposed, solutions which were once considered inferior can now be considered noninferior. This argument leads to a natural conclusion: since it is probably impossible to elicit all objectives from groups of decision makers, it is important to be able to generate both noninferior solutions and close to noninferior solutions. Techniques that support collaborative decision making must be capable of generating close-to-optimal alternatives, of searching for good compromise solutions, and of searching for solutions that differ spatially but are not very different in performance.

Collaborative decision making involves generating feasible alternatives among many individuals or groups. It is often difficult to formulate problems to include feasibility factors such as political aspects, human perceptions, safety factors, aesthetics, etc. Some process of visualizing, evaluating, and adjusting model generated alternatives is required to develop a feasible group consensus. Techniques need to be developed to intelligently explore the decision space of spatial problems and to look for good (feasible) solutions to ill-defined problems.

6.1.4 Revealing preferences and objectives

Economists often infer the relative value of various objectives of a decision maker by determining which weights yield an optimal choice similar to that made by the decision maker, or by asking a decision maker to choose between a series of pairwise comparisons. Understanding which objectives are important, whether voiced or not, can be important in reaching an accord. Clearly, systems which can help identify underlying preferences or objectives can aid collaboration and negotiation.

Consider the following example: suppose a decision maker had selected a specific route for a highway alignment. According to an analysis based on tradeoff of objectives, it is clear that the decision maker is interested in ensuring that a specific town is close to the route. Using this information, it is then possible to generate tradeoffs in the route selection based on total vehicle miles traveled by others vs. the total vehicle miles traveled by people in this specific town. The decision maker may then see the cost of meeting his desired goal (getting close to a specific town) as a function of the cost to all others. Without identifying what objectives are present or the relative importance of those objectives, it may be impossible to tease entirely rational designs or negotiate a best compromise in a collaborative decision making setting. An important research objective is to look at alternatives for capturing decisions and revealing preferences in spatial problems, and to test various approaches in prototypes.

6.1.5 Problems of presenting multiple solutions and visualizing differences

The presentation and comparison of alternative solutions in many spatial decision support systems is poorly conceived at best. Few examples exist where the interface design had an emphasis on the presentation of differences between alternative solutions. Thus, not only is it important to be able to study a given solution, but also to be able to spatially compare different solutions in terms of both objective and decision space attributes. Example designs and prototypes should be developed to test approaches which might be useful to accomplish this task.

6.1.6 Using animation to examine sensitivity to change and to examine change over time

Animation can provide a tool for viewing how a solution changes as a result of changing model parameters. After a model is solved, it is often important to understand how sensitive a given solution is to the original model parameters. Often this is done by systematically changing the model's parameters to see if changes result in the same solutions--a process which can be very time consuming and produce results which are difficult to compare. Currently, for most spatial optimization models, there is no automatic way in which to generate and view such demonstrations of model sensitivity. Animating sensitivity analysis can aid in the understanding of input data error and uncertainty, and may allow complex spatial models and their solutions to be viewed in a form which may help reveal specific nuances (e.g. why is this area never chosen).

Given that some model solutions are temporal as well (spanning up to 20 decades), animation may also be an important tool for viewing how a solution changes over time. Insight into temporal change may provide some important common ground for a group of decision makers who are considering a number of different solutions.

To address some of these research themes, the following research is planned at UCSB:

  1. Identify and formalize the geographic conceptions of problems inherent in the interests of participants in multiparticipant decision situations. This may use a variety of methodologies including experimental techniques, ethnographic techniques and other methods for analyzing text or discourse. Initially, research will focus on land use debates because they pose the biggest problem in terms of divergence in the conceptions of the problem. Content analysis techniques will be used on records of a land use debate to identify the geographic concepts that are important in this debate. Analysis of experimentally derived protocols may provide additional data.
  2. Analyze common representations of spatial information that are used in SDSS to determine their efficacy in addressing the interests of participants in debates. This will require formalizing the types of information that are explicit or implicit in representations of spatial information. These formalizations will be compared to the concepts identified in the analysis of the land use debate (this will use a knowledge representation language such as Conceptual Graphs) and tested for their ability to represent the participants' interests. These analyses should lead to the identification of needed extensions to common representations.
  3. Develop a prototype for generating spatial alternatives in a spatial decision support system. The major objective will be to develop and test a method for generating alternatives, a graphical user interface to present alternatives, and a method to spatially direct searches for feasible alternatives. Such a tool can be used in collaborative situations to provide comparison and examination of similar solutions.

6.2 Related research at State University of New York at Buffalo

The following work is proposed by Marc Armstrong (University of Iowa) and Paul Densham (University College London, UK) who are working in collaboration with the NCGIA at SUNY Buffalo.

6.2.1 The cartography of collaboration

Collaborative spatial decision-making environments in which group members individually and collectively pursue solutions to ill-structured problems have a unique set of cartographic visualization requirements. Group members normally have varying levels of education, disciplinary backgrounds, and familiarity with computing, as well as different stakes in, and degrees of familiarity with, ill-structured problems. Consequently, we can expect that group members will articulate different types of questions and will have considerably different perspectives on the way that these questions should be addressed. The purpose of our work is to develop a cartographic framework that supports the design, construction and use of maps in CSDM. The central principle in this framework is that each map created by an individual as part of a solution to a problem can be decomposed into a collection of atomic objects - a path through a network can be decomposed into a series of nodes and links, for example. These objects are then placed into an accounting framework that supports summary operations on the objects and enables group members to determine the level of agreement among geographically- distributed components of alternative solutions.

6.2.2 The role of intelligent agents in CSDM

The range of tasks and types of applications that need to be supported in CSDM environments is characterized by great diversity, since they are often constructed from a number of different software modules. This interoperability problem in CSDM is a difficult one to treat, however, because great differences exist among the user interfaces of software modules and each module typically has unique data flow requirements. Software agents represent one attempt to circumvent such interoperability problems. Agents also may actively assist users who may be unfamiliar with the operation of software. The purpose of this work is to articulate a vision of how agents can be used to support decision-makers and to develop a conceptual framework for the roles of agent- based computing in CSDM environments.

6.2.3 Visual interactive modeling

The lack of structure inherent in many complex spatial problems makes it difficult for individuals to understand the relationships among different components of a problem. Consequently, individuals require tools that help them to explore and understand problems as well as resolve them. In many settings, human-computer interaction is enhanced if each user can articulate their ideas by interacting directly with graphical representations of their problem. When faced with a decision about where to locate a school, for example, users could drag the symbol for the school to different locations on a map and watch the system enumerate and display in real time the concomitant changes in enrollment, age structure, gender and ethnic ratios, and distances traveled; an alternative approach is to specify some criteria for selecting a location and invoke an optimizing spatial search procedure. In such a context, a visual interactive modeling environment provides analytical capabilities that are invoked using map windows and linked tabular views that help groups of decision-makers to understand and reconcile depictions of spatial pattern with statistical reports about locational configurations. The purpose of the work proposed under this heading is to take a fresh look at the design, representation and implementation of spatial models. More specifically, we intend to extend earlier work on the design and implementation of modelbase management systems (MBMSs) into the domain of CSDM to meet the challenge of providing flexible modeling tools for group use. We will build substantially upon research carried out under I- 6 (Spatial Decision Support Systems) and that described above on the cartography of collaboration and the role of intelligent agents.

   Go forward to Appendix A

   Go back to Table of Contents

Posted March 28, 1996

Comments to Karen Kemp